The online optimization of optical microscopy parameters aims at learning the set of imaging parameters while the experiment unfolds. The success of this task relies on the trade-off between multiple confounding objectives and may vary depending on the experimental setting. In this work, we frame the optimization problem as a multi-armed bandit framework with contextual information about the sample to identify optimal sample-dependant imaging parameters. This allows to take into consideration the current state of the sample and choose the imaging parameters accordingly.
Article ID: 2022S01
Publisher: Canadian Artificial Intelligence Association